Comparison of PLSR modelling and indicative mineral mapping of airborne hyperspectral imagery for acid sulphate soil assessment

Abstract

Acid sulphate soils (ASS) are widely distributed around the world and can be harmful to the environment due to their potential release of severe acidity, which in turn can mobilize harmful quantities of both major and trace metals. The effective mapping and assessment of ASS and the resulting spread of their harmful effects are important in the management of these widespread soils. Secondary iron and sulphate-bearing minerals form within and on surfaces of AAS during oxidative evolution. These secondary minerals are indicative of the existence of different pH conditions on the surface of the soil. Many of these indicator secondary minerals associated with acidic soil conditions can be identified by hyperspectral sensing due to their diagnostic spectral reflectance features. Accordingly, hyperspectral sensing was used in a coastal area bearing AAS to identify and map secondary minerals using spectral absorption features. This information was used to establish the spatial extent and severity of soil acidity by utilizing the relationship between the presence of indicative minerals and pH values. Additionally, an intrinsic relationship between pH values and reflectance spectral features was also modelled by the partial least square regression (PLSR) method using ASS samples collected from the test site and applied to HyMap imagery to successfully deduce an acidity map. Both resultant maps of acidic conditions were compared and it was found that nearly 94% of the pixels in the two pH maps deduced from these different methods were highly similar. This suggested that the soil pH distribution attained was accurately mapped by the HyMap imagery and the PLSR model established was robust at predicting soil acidity affected by ASS in the study area.